Meta Data Scientist Product Analytics Interview Guide

This guide uses the internal Meta DS interview docs compiled by their own data scientists as a bible (which we have thanks to our coaches). The round names and focus areas for each in this guide are verbatim and also feature advice from Prepfully’s coach network, including Staff and Lead DSs. 

The interview loop officially goes through these stages:

1. Recruiter screen
2. Technical Screen
3. Onsite interviews:
   Technical Skills
   Analytical Execution
   Analytical Reasoning
   Behavioral
4. Hiring Committee and Team Match

This is a 30 minute call on which you'll talk to your recruiter about your

• background
• experience
• long-term goals, 
• why Meta 

and other relevant details so they can lock in a level that sets the bar for how you’ll be evaluated for the rest of the loop.

They'll also ask you about your areas of expertise and experience working with relevant programming languages and tools. Expect a few questions on how you've addressed technical challenges in some of your previous projects.

There are 2 things to do on this call:

• Inform the recruiter what language you'll use for the technical interviews. Nearly every Product Analytics candidate goes with SQL.
• Establish a good rapport with your recruiter by being upfront about your background and what you want out of your career. They’re known to advocate for a candidate later in the process so they're good people to have on your side. 

This is a 45 minute interview that combines technical coding with a case study, with a few minutes at the end for your questions. You'll typically get one or more table descriptions with some sample rows, and be expected to write code in CodePad. The code won’t be executed. 

The round exists to assess your ability to query and manipulate data, build analysis plans, design experiments around new features, and define success metrics for them. 

The technical screen scores Programming, Determining Goals & Success Metrics, Data Analysis and Research Design. 

For Programming, Meta says the assessment centers on common data manipulation, namely dataset merging, filtering data for insights and handling missing data.

Determining Goals & Success Metrics is assessed in two types of questions:

• Metric definition: selecting a success metric
• Metric Analysis: where you’re asked to investigate a movement, typically framed as X moving Y% in the last N days 

Data Analysis focuses on how you answer questions that are explorative/ hypothesis based to come up with a plan.

Research design covers how you meticulously design experiments and evaluate them. You should be able to tie the design back to business outcomes and think through any downstream effects that the experiment should correct for.

Candidate reports agree on the ingredients and disagree about the proportions. Some reports range from a single SQL question in two parts followed by a product case and some report 5-6 SQL questions on one business case followed by a full A/B research design.

Expect questions like:

• Given a user activity table, find the number of users' interactions since the user first logged in to Facebook
• How would you determine the health of Facebook Groups?
• An A/B test confirmed a 0.2% drop in CTR but an increase in avg session length. How would you interpret this? 
• What's the difference between a left join, a union, and a right join?

The entire loop is 4x 45 minute interviews:

1. Technical Skills
2. Analytical Execution
3. Analytical Reasoning
4. Behavioral

As for interviewers, you’ll meet future peers, cross functional partners and other leaders.

Once you reach the stage, the Meta Connections program lets you talk with an employee about what working at Meta is like. If you want to get a headstart and choose the peer you want to talk to, you’ll find a list of Meta DS across seniorities here. 

1. Technical Skills 

This is the coding interview. Meta frames its questions as open-ended data problems where you'll 

• structure and articulate a solution based on the data 
• code an executable version of that approach
• identify and handle any edge cases 
• adapt your solution around new data or constraints the interviewer introduces (usually after you have a working baseline)
• stay mindful of efficiency

Officially, this round will be scoring Programming, Communicating Effectively, Data Analysis and Determining Goals & Success Metrics.

Since Research Design from the screen gets swapped for Communicating Effectively, we must spend a second to understand what this means. It’s how you give people timely and concise information while adjusting it for the people who are receiving it. 

Expect questions like:

• Find the number of users who called at least 3 people in the last week from a table that logs peer-to-peer calls.
• Write a function to check if a string of user text is a valid palindrome
• Given an event log of user logins, identify which user demographic had the largest drop in activity between yesterday and today. 
• How do you create a validation tool for Facebook Marketplace?

Our coaches also mention:

• Any convenient magic functions that’ll trivially solve the problem are fully discouraged.
• Of course the interviewers will know widely used functions and syntax, but you should go ahead and explain what you're reaching for just in case the interviewer isn’t familiar.
• Take the effort to write your code out in full. Meta draws the line at pseudocode that skips details.

To be thorough with complex data extraction and query optimization across massive datasets, refer to our Amazon Data Scientist Guide and Walmart Data Scientist Guide. They cover real-world ETL and DB management scenarios.

2. Analytical Execution

This is the loop’s statistics and hypothesis section. You’ll be handed an open-ended and ambiguous product/ business scenario and asked to make decisions using data. 

Now, the aspects that Meta lists for this round fall into 2 groups: 

• Product: that covers your hypotheses around launching new features and considering and quantifying a feature's tradeoffs in terms of metrics
• Statistics: descriptive statistics (mean or expected value, median, mode and percentiles), common distributions such as binomial and normal, the profile of real-world data, the Law of Large Numbers, the Central Limit Theorem, linear regression, and conditional probability including Bayes' Theorem.

You’ll be scored on the following official areas:

• Creating Hypotheses: the quality of your assertions and how you form them to test ideas and answer business questions relevant to the prompt
• Quantitative Analysis: how well you use statistical code for analysis ranging from correlations to multivariate analysis and measurement models.
• Goals & Success Metrics
• Demonstrating Agility: how you proactively and flexibly adopt change like new data/constraints while still tackling ambiguity  

A note on what won’t be tested: 

• calculus 
• advanced statistical/ ML models 
• complex distributions (Weibull, Beta, etc) 
• contrived estimation problems 

Sarayu Sarangan writes in Meta’s official blog about the “Ladder of evidence in understanding effectiveness of new products,” for when standard A/B tests aren't feasible when launching to 100% of users or dealing with high network spillover. These are scenarios where you’re expected to deploy Pure Quasi-Experimentation that you can read about in the blog. 

Prepfully candidates who recently appeared for this round reported facing questions around  evaluating a new feature, diagnosing a drop in user engagement, or managing metric tradeoffs. 

• If we double the number of ads shown on Instagram, what network effects should be kept in mind while designing the A/B test? How do you prevent spillover?
• An A/B test shows a 0.2% drop in CTR but an increase in session length. What is your launch recommendation?
• Explain the difference between a t-test and z-test. When would you choose one over the other?

3. Analytical Reasoning

This is a product case interview that’s conversational and features no dataset. You’ll be expected to reason through a high-level product or system situation and account for a whole bunch of complex dynamics like causal inference, selection bias, seasonality, and network effects using pure judgment. 

As you talk through your solution, the interviewer will once again introduce changes or new signals to see if your logic still stands. 

This round officially scores: 

• Research Design 
• Analytical Design: the quality of your analysis plan that sets out the statistical software and techniques needed to arrive at  conclusions from the data you’re given. 
• Storytelling Through Data: how you help both technical and non-technical peers understand everything that you do: how you solved a business problem and the tradeoffs made
• Setting Goals & Success Metrics: going a little further from the other rounds, includes sizing opportunities and setting goals starting at the project level, all the way up to the business level
• Data Visualization: how insights are depicted visually for trends, outliers, patterns, etc. 

Recently, candidates have their discussion kicking off with:

• We’re thinking of launching a feature where you’re notified when a friend attends an event. Should we do it? 
• Friend request acceptance plummeted by 15% after a notification system went live. How would you investigate this?

Apart from the prompts, expect questions that touch upon the process of the round like:

• Your approach to framing ambiguous product scenarios
• What's important while designing experiments to test hypotheses
• How to pick out the datasets best suited for your task
• How to surface the downsides of a proposed methodology and identifying the biases in the analysis/experiment 
• How to draw inferences from a data set/ combine multiple signals into a statement
• How analytical insights map back to product impact
• How you help teams make decisions by choosing the right metrics

Some tips for this round: 

• Hold off on AB testing till the metrics and the reason for the change are settled. Generally speaking, for any of the technical interviews that involve ambiguous problems, you should keep checking in with your interviewer, asking them simple questions like, “is this how you'd like me to proceed?”. 
• Data visualization isn't scored anywhere else in the loop except for this round. So you should already have a view of how you would like the result to be shown after you end on a recommendation you’re willing to defend. 

4. Behavioral

This round is focused on evaluating your alignment with Meta's core values while discussing professional challenges and your scope in operating those. The questions usually take the shape of “tell me about a time when” to explore relationships you have with your stakeholders, how you handle disagreements or how you’ve led a data project. The questions mostly lean on your previous experience but be prepared for hypothetical ones too. 

Behavioral scores 5 focus areas: 

• Growing Continuously: actively seeking out and valuing chances to grow and learn professionally 
• Partnering: building a diverse network of relationships that you can influence without leaning on authority/rank to achieve shared goals
• Building Inclusion: seeking and valuing different perspectives and feedback without getting defensive or trying to “win” 

The other two, Demonstrating Agility and Communicating Effectively carry over from the technical rounds which hints at some light behavioral-flavored questioning in the other rounds too. 

Expect questions like: 

• What do you do if someone from a different org challenges your approach?
• Tell me about a time you used data to clarify something when everybody else was working off instincts.
• What has been your most complex project so far?
• How do you keep your xfn partners (SWEs, PMs, etc) accountable for all their deliverables when you’re working on a project?

To have an anecdote for every type of scenario or follow up, we suggest that you make a story bank. Make sure you have stories about:

• Operating in ambiguous projects/ taking the lead on undefined ones
• Moving quickly and while showing resourcefulness and a bias for action
• Transparency on failures and what you learnt from them
• Building relationships with direct and partnering teammates to work together 
• Convincing/ getting buy-in from peers who initially resisted your goals/ plans
• Self awareness and introspection wrt your professional growth and personal impact thus far

We recommend that you take the classic STAR framework and build it out into STAR+ (Situation, Task, Action, Result + the X-Factor: Learning) to structure your stories. The X-Factor at the end - Learning - is your definitive chance to demonstrate intellectual humility, and helps you reflect on a personal takeaway on how that situation changed you for the better. 

At the end of each round, interviewers get 24 hours to submit their notes and pick between Strong Hire, Hire, Leaning No Hire, or No Hire for you. 

The Hiring Committee, consisting of senior DSs then reviews that feedback to make the final call and decide your level. 

If the committee signs off on a “yes”, you’ll then move into the Team Matching phase where you'll do 30-minute informal chats with HMs until you zero in on a fit.

IC3: $165K

IC4: $249K

IC5: $351K

IC6: $564K

View the latest compensation for Meta DSs on Levels.fyi